Add SSL and JSSL functionality

direct/config/defaults.py

@@ -1,4 +1,3 @@
-# coding=utf-8
 # Copyright (c) DIRECT Contributors
 
 from dataclasses import dataclass, field
@@ -102,6 +101,7 @@ class InferenceConfig(BaseConfig):
 @dataclass
 class ModelConfig(BaseConfig):
     model_name: str = MISSING
+    engine_name: Optional[str] = None
 
 
 @dataclass

direct/data/datasets_config.py

@@ -1,4 +1,3 @@
-# coding=utf-8
 # Copyright (c) DIRECT Contributors
 
 """Classes holding the typed configurations for the datasets."""
@@ -10,6 +9,14 @@
 
 from direct.common.subsample_config import MaskingConfig
 from direct.config.defaults import BaseConfig
+from direct.data.mri_transforms import (
+    HalfSplitType,
+    MaskSplitterType,
+    RandomFlipType,
+    ReconstructionType,
+    SensitivityMapType,
+    TransformsType,
+)
 
 
 @dataclass
@@ -22,7 +29,7 @@ class CropTransformConfig(BaseConfig):
 @dataclass
 class SensitivityMapEstimationTransformConfig(BaseConfig):
     estimate_sensitivity_maps: bool = True
-    sensitivity_maps_type: str = "rss_estimate"
+    sensitivity_maps_type: SensitivityMapType = SensitivityMapType.RSS_ESTIMATE
     sensitivity_maps_espirit_threshold: Optional[float] = 0.05
     sensitivity_maps_espirit_kernel_size: Optional[int] = 6
     sensitivity_maps_espirit_crop: Optional[float] = 0.95
@@ -34,7 +41,7 @@ class SensitivityMapEstimationTransformConfig(BaseConfig):
 class RandomAugmentationTransformsConfig(BaseConfig):
     random_rotation_degrees: Tuple[int, ...] = (-90, 90)
     random_rotation_probability: float = 0.0
-    random_flip_type: Optional[str] = "random"
+    random_flip_type: Optional[RandomFlipType] = RandomFlipType.RANDOM
     random_flip_probability: float = 0.0
     random_reverse_probability: float = 0.0
 
@@ -47,6 +54,61 @@ class NormalizationTransformConfig(BaseConfig):
 
 @dataclass
 class TransformsConfig(BaseConfig):
+    """Configuration for the transforms.
+
+    Attributes
+    ----------
+    masking : MaskingConfig
+        Configuration for the masking.
+    cropping : CropTransformConfig
+        Configuration for the cropping.
+    random_augmentations : RandomAugmentationTransformsConfig
+        Configuration for the random augmentations.
+    padding_eps : float
+        Padding epsilon. Default is 0.001.
+    estimate_body_coil_image : bool
+        Estimate body coil image. Default is False.
+    sensitivity_map_estimation : SensitivityMapEstimationTransformConfig
+        Configuration for the sensitivity map estimation.
+    normalization : NormalizationTransformConfig
+        Configuration for the normalization.
+    delete_acs_mask : bool
+        Delete ACS mask after its use. Default is True.
+    delete_kspace : bool
+        Delete k-space after its use. This should be set to False if the k-space is needed for the loss computation.
+        Default is True.
+    image_recon_type : ReconstructionType
+        Image reconstruction type. Default is ReconstructionType.RSS.
+    pad_coils : int, optional
+        Pad coils. Default is None.
+    use_seed : bool
+        Use seed for the transforms. Typically this should be set to True for reproducibility (e.g. inference),
+        and False for training. Default is True.
+    transforms_type : TransformsType
+        Type of transforms.  By default the transforms are set for supervised learning (`TransformsType.SUPERVISED`).
+        To use SSL transforms, set transforms_type to `SSL_SSDU`. This will require additional parameters to be set:
+        mask_split_ratio, mask_split_acs_region, mask_split_keep_acs, mask_split_type, mask_split_gaussian_std.
+        Default is `TransformsType.SUPERVISED`.
+    mask_split_ratio : Tuple[float, ...]
+        Ratio of the mask to split into input and target mask. Ignored if transforms_type is not `SSL_SSDU`.
+        Default is (0.4,).
+    mask_split_acs_region : Tuple[int, int]
+        Region of the ACS k-space to keep in the input mask. Ignored if transforms_type is not `SSL_SSDU`.
+        Default is (0, 0).
+    mask_split_keep_acs : bool, optional
+        Keep ACS in both masks, input and target. Ignored if transforms_type is not `SSL_SSDU`. Default is False.
+    mask_split_type : MaskSplitterType
+        Type of mask splitting if transforms_type is `SSL_SSDU`. Ignored if transforms_type is not SSL_SSDU.
+        Default is `MaskSplitterType.GAUSSIAN`.
+    mask_split_gaussian_std : float
+        Standard deviation of the Gaussian mask splitter. Ignored if mask_split_type is not `MaskSplitterType.GAUSSIAN`.
+        Ignored if transforms_type is not `SSL_SSDU`. Default is 3.0.
+    mask_split_half_direction : HalfSplitType
+        Direction to split the mask if mask_split_type is `MaskSplitterType.HALF`.
+        Ignored if MaskSplitterType is not `HALF` or transforms_type is not `SSL_SSDU`.
+        Default is `HalfSplitType.VERTICAL`.
+    """
+
     masking: Optional[MaskingConfig] = MaskingConfig()
     cropping: CropTransformConfig = CropTransformConfig()
     random_augmentations: RandomAugmentationTransformsConfig = RandomAugmentationTransformsConfig()
@@ -56,9 +118,17 @@ class TransformsConfig(BaseConfig):
     normalization: NormalizationTransformConfig = NormalizationTransformConfig()
     delete_acs_mask: bool = True
     delete_kspace: bool = True
-    image_recon_type: str = "rss"
+    image_recon_type: ReconstructionType = ReconstructionType.RSS
     pad_coils: Optional[int] = None
     use_seed: bool = True
+    transforms_type: TransformsType = TransformsType.SUPERVISED
+    # Next attributes are for the mask splitter in case of transforms_type is set to SSL_SSDU
+    mask_split_ratio: Tuple[float, ...] = (0.4,)
+    mask_split_acs_region: Tuple[int, int] = (0, 0)
+    mask_split_keep_acs: Optional[bool] = False
+    mask_split_type: MaskSplitterType = MaskSplitterType.GAUSSIAN
+    mask_split_gaussian_std: float = 3.0
+    mask_split_half_direction: HalfSplitType = HalfSplitType.VERTICAL
 
 
 @dataclass